Objective Rating of Seam Pucker using Neural Networks

Abstract

An objective method of evaluating seam pucker in woven fabrics during garment manufacturing is studied using artificial neural networks. An automatic sewing machine and new measurement system with a laser sensor are presented. For objective evaluation of seam pucker using AATCC standards, two artificial neural networks are constructed from pattern recognition and learning. An error backpropagation model is adopted for the neural networks. The puckered shape of a sewn fabric is converted into the numerical data on three-dimensional coordinates by the laser scanning system. Measurement data in a parallel direction with the seam line are transformed into power spectra on the frequency domain using fast Fourier transformation. The power spectra then generate the specified patterns for neural networks. Finally, the neural networks evaluate seam pucker the same way as the AATCC rating of well trained human experts.